metadata
dataset_info:
features:
- name: country
dtype: string
- name: capital
dtype: string
splits:
- name: train
num_bytes: 4776
num_examples: 197
download_size: 5587
dataset_size: 4776
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
A benchmark dataset for slms. Contains every country and capital pair. The capitals.yaml file contains the config file for use in lm-evaluation-harness.
